👁️ Cataract Detection & Severity Prediction

Deep Learning powered system for cataract classification and severity estimation
Built with PyTorch, Torchvision, and ResNet architectures.

---

📌 Overview

This repository contains two complementary modules:

• main.py – Trains a ResNet50 model for severity prediction (regression output as percentage).
• test.py – Loads a ResNet18 model for binary classification (Cataract vs No Cataract) and predicts on new images.

---

✨ Features

• 📊 Severity Prediction – Outputs cataract severity as a percentage using regression (MSE loss).
• 🩺 Binary Classification – Distinguishes between Cataract and No Cataract cases.
• ⚡ Transfer Learning – Fine-tunes pre-trained ResNet models for medical imaging tasks.
• 🎯 Data Augmentation – Includes resizing, rotation, color jitter, and normalization for robust training.
• 💾 Model Persistence – Saves trained weights for later inference.

---

🛠️ Tech Stack

• Frameworks: PyTorch, Torchvision
• Models: ResNet50 (regression), ResNet18 (classification)
• Tools: Matplotlib, PIL

---

⚙️ Usage

Training (Severity Prediction)

python main.py

This will:

• Train ResNet50 on cataract images
• Save weights to cataract_severity_model.pth
• Report mean severity prediction error

Testing (Classification)

python test.py

This will:

• Load ResNet18 with trained weights (cataract_model.pth)
• Predict Cataract vs No Cataract for a given image
• Print the predicted class

---

📂 Repository Structure

├── main.py        # Train ResNet50 for severity regression
├── test.py        # Test ResNet18 for binary classification
├── Cataract/
│   └── processed_images/
│       ├── train/
│       └── test/
└── README.md